<OL> <LI> Develop an aflatoxin control program for the U.S. almond industry by (a) determining which almond grade components can be considered aflatoxin risk components, (b) evaluating the effect of electronic color and hand sorting methods on reducing aflatoxin in raw shelled almonds, and (c) from existing variability and distributional data, design and evaluate the performance of aflatoxin sampling plans for exports lots destined for the EU. <LI> Develop sampling plan to detect Cry9c protein in export grain by (a) determining the Cry9c protein distributional among individual corn kernels, (b) determining the variability associated with sample preparation (grinding and subsampling), and analytical step of the test procedure used to measure Cry9c protein in shelled corn, and (c) developing a statistical model that use variability and distributional information to evaluate the performance of sampling plans that measure Cry9c protein in bulk lots of shelled corn. <LI> Develop sampling plans to detect aflatoxin and ochratoxin A in dietary supplements by (a) measuring the variability and distribution among sample test results when sampling retail lots of powdered ginger for aflatoxin and ochratoxin A; and (b) developing statistical models to evaluate the performance of sampling plans that measure aflatoxin and ochratoxin A in powdered ginger lots using the variability and distributional data.
NON-TECHNICAL SUMMARY: Once the producer markets an agricultural product, processors use different methods to increase the value of the lot by removing unwanted attributes such as genetically modified seed and mycotoxin-contaminated kernels from the lot. For nut products, many of the sorting methods used in the processing steps remove contaminated kernels from the lot. Often nut categories (grades) that are considered high mycotoxin risks are not known and the efficiency of sorting methods at removing high-risk nut categories, which will reduce lot aflatoxin concentration, is not known. After lots are processed, sampling plans (defined by sample size, sample preparation methods, and analytical methods and a defined tolerance) are used to detect unwanted characteristics (mycotoxins and genetically modified seed) in a lot so they can be diverted from food use. Because of the variability among sample test results taken from a bulk lot, the true level of a characteristic in the lot cannot be determined with 100% certainty. Because of this variability, some lots will be misclassified by the sampling plan. Some good lot will be rejected (false positive or seller's risk) and some bad lots will be accepted (false negatives or buyer's risk) by the sampling plans used to measure the level of an attribute in agricultural products. Buyers and sellers of a product want to reduce both these risks to the lowest possible level because of the high costs involved in misclassifying lots. Industry representatives have defined research needed to improve food safety in the following areas. <P>
APPROACH: <BR>Objective 1. From each of 50 lots of shelled almonds, a 12 kg samples will be randomly selected and graded by qualified inspectors. Each 12 kg sample of almonds will be divided completely into five grade components and analyzed for aflatoxin. Five lots of shelled almonds will each be sorted sequentially first by laser sorters and secondly by human sorters in handler facilities in California. The accept stream from the laser sorter will then be hand picked and partitioned into reject and accept streams. Samples will be randomly taken from each of the following five stream locations and identified by lot number and stream location, weighed, and analyzed for aflatoxin. Because of EU revisions, the almond industry will also revise its aflatoxin-sampling program for lots exported to the EU. The maximum risk associated with accepting bad lots and rejecting good lots will be defined by the industry. The exporter's risk and importer's risk associated with various aflatoxin sampling plan designs will be computed with a statistical model developed from previous sampling studies. The performance of each aflatoxin sampling plan design will be provided to the almond industry. <P>Objective 2. The variability and distribution associated with sampling, sample preparation, and analytical steps used to measure Cry9c in shelled corn must be determined in order to develop a statistical evaluation model. For the sampling step, the Cry9c distribution among 300 corn kernels will be determined. Sampling lots with a given concentration using samples of various sizes will be simulated using Monte Carlo methods. The variability associated with sample preparation and analysis will be determined using a balanced nested design. The subsampling and analytical variance components will be computed. Functional relationships between subsampling and analytical variances and Cry9c concentration will be determined. Monte Carlo methods will be used to determine the performance of various designs can be predicted. <P>Objective 3. The variability and distribution associated with sampling and analytical steps used to measure aflatoxin (AF) and ochratoxin A (OTA) in ginger will be determined. Twenty bottles of ginger will be randomly selected from each of three retail lots. From each bottle, four samples will be randomly selected. From each sample, AF and OTA will be extracted and two aliquots will be removed from the extract and quantified for AF and OTA. The total variance will be partitioned into sampling and analytical variances for AF and OTA. The distribution among sample test results will be compared to the normal and lognormal to find a model to simulate the distribution among sample test results for AF and OTA. Because ginger is also sold in bags, the same process as described for bottles will be repeated for ginger sold in bags. From the variability and distribution information, two statistical models will be developed to predict the performance of sample size, number of aliquots quantified, and accept/reject limits on the performance of AF and OTA sampling plans. The performance of various designs will be shared with FDA officials.